0198-0149/84 52.00 + 0,W Q 1984 Pcrgrmon Prar Ltd.

Meritdional and seasonal variations of the sea [surface chlorophyll concentration in the southwestern tropical Pacific (14 to 32OS, 160 to

175 OE)

YVES DANDONNEAU* and FRANCIS G0lilN"t

(Recelved 5 Noivnrbcr 1983: acccpred I6 Febnrarjl 1984)

Abstrrcl-Sea xurf'ace chlorophyll coircentrations (SSCC) (3369) have hccn measured from Jnnuary IY78 to December 1982 in the arca 14 to 3 2 O S . 160 to 175OE, on samples taken by vufuntnry observing ships. A 1og.transf'ormation is first applicd lo rhc data to stnbilize the variance: u furiction of time and lntitudc is obtained which accounts Tor IbYh of the variance of' the logs. I t is lhcn subtracted from the logs. resulting in a field with a zcro mean and,'Ú constant and uncorrclntcd crror. In which the longitude cffcct is small and will bc neglcctcd. Thc dbta nrc distributed into a onc- nionth, twodegree latitude grid: the subgrid noise variuncc is 0.472 aQd corresponds to n variation ccwficirnt of 711% arter return to the antilog. Objcctivc analysis ir appl.ied to obtain optimal estiniates of the SSCC in the limits of' the study. Thc nlost slriking feature is an enrichment in austpd winter. dcveloping northward to about 20 to 22'5, with SSCC in the range 0.15 to 0.40 mg m-'; A poritivc c'jrrclation bctwecn SSCC and inlegrricd (O to 150 m) chlorophyll (log transformfition. r 1a0.714. II u 150) is round in the samc region using existing data rrom occnnographiq cruises: thc rcrulta from SSCC data can thus cautiously be applied to the wholc euphotic Inyer. The data from crcanogrnphic cruises support the hypothesis that thc winler enrichment results from nn cnhanccd vcrticnl mixing arter the cooling or thc sea surfncc. This is confirmed by the direct rcla[ionship hctwecn the f'requcncy of' strong winds and thc importancc of' the winter enrichment. Thcrc is Some indication thal the coinbina!ion of avnilablc light nnd mixed-laycr depth is limiting south of 28's in Junc nnd July. resulting in two mnxinia in May and August. The wigd effect is invcrse in the northern part where Trlcliodesmirrm spp. patches form during the calms: lhesc patches nrc only surfncc features and arc not indicative of a biomass increase at deeper levels. The interannual urriability is rcflccted by the ratio of'thc range of the standard year to that or the monthly anomalies which varies from 0.5 nt lSoS to 1.1 I at 29OS.

I

> ,

INTRODUCTION

IN temperate seas, the primary production and other phytoplankton related parameters exhibit seasonal cycles which are strongly dependent upon the wintcr cooling and vertical mixing of the surface layers. CUSHING (1959) proposed that the time lags between the winter supply of nutrients. the phytoplankton bloom, and the zooplankton maximum, decrease with latitude. and that the ratios between the extreme values of algal or herbivore production.(i,e., the annual signal) also decrease from polar to tropical regions, A similar scheme was proposed by BOGOROV ( I 958). This annual signal is well documented in temperate latitudes where i t has a large amplitude, while in low latitudes fewer studies have often failed to describe n vnnishing annual signal. Reviewing lhe question of annual cycles of phytoplankton in *, .

* Gmupc SURTROPAC, Centre ORSTOM, B.P. A5 Noum6a, New Caledonia. t Present address: Antenne ORSTOM. Centre Ocinnologique de Bretagne, B.P. 337, 29273 Ercsl. Francc.

' 1377

---A-

P

b 1

i

8

.' _I ;

1378 Y. DANDONNEAU and F. GOHIN

tropical seas, SOURNIA (1969) observed that in oceanic areas the information is too scarce and inadequate to substantiate the existence and modalities of annual cycles at low latitudes: he also concluded that ecological factors other than the winter vertical mixing might be deter- minant in tropical seas. A later study, by OWEN and ZEITSCHELL (19701, concluded that seasonal variations are significant in the eastern Pacific; their data (from six cruise periods) present a two-fold variation over one year in an area where a thermal dome, an equatorial divergence, a coastal upwelling, and river outflow all have more or less independent dynamic signals. Observing a three-fold variation in the chlorophyll content of the upper 260 m (and a 26-fold variation in the carbon fixation per square meter!), BIENFANG and SZYPER ( I 98 1) con- clude that there is "no evidence of seasonality" off Hawaii, These two last studies poorly describe long-term variations (the signal) since the available data do not allow an estimate o f the short-term variability (the noise): in both cases, six cruises are nyt sufficient to account for seasonal variations.

Using measurements at two-week intervals over a period of three years at a fixed point near Bermuda (32ON), MENZEL and RYTHER ( I 961) clearly illustrated an enrichment period which is related to the winter cooling of the mixed layer, and to the increased vertical turbulence. At lower latitudes (Jamaica, 18ON, and Barbados, 15ON), with monthly observations over a two-year period, BEERS et al. (1968) found more uncertain seasonal variations; the uncertainty comes both from a smaller seasongl.signal, and from a loose network of observa- Lions, However, the decrease of the seasonal signal with decreasing latitude is emphasized by these two studies which were carried out in 'neighbouring areas.

The purpose of the present study is to describe the changes in seasonality with lntitudc in thc southwestern tropical Pacific, Such a view requires a large amount of data adequatcly distributed both in time and space, which is seldom gathered in biological oceanography. though several large-scale studies of primary production have been based on oxygen data (MINAS, 1970; VOITURIEZ and HERBLAND, 198 1 ; SCHULENEERGER and REID, I98 I ) .

From 1978 to 1982, 3369 sea surface chlorophyll concentrations (SSCC) have bcen measured along the routes of merchant ships between 32 and 1 4 O S , 160 and 1 7 5 O E (Fig. I ) .

10's

I164 1 7 54 81 L6

30's

14O'E 166. E 180' Fig. 1. Numbcrs of observations.

L

i

SSCC in the tropical Pacific

[f

1379

as part of thc SURTROPAC program of the Centre ORSTOM dc Noumia (New Caledonia); thcy constitute the main source of information analyzed here; the significance of the SSCC concerning the whole euphotic layer is discussed later. As this data set is very noisy, and prc- sents some gaps, a technique of optimal objective mapping has beenmed (BRETI-IERTON et al., 1976: KARWEIT, 1980); a section is devoted to the problems raised:by the application of this iechniq ue.

MATERIALS A N D METHODS I. 5

The mcthod used to Òbtain the chlorophyll concentration values has already been described (DANDONNEAU, 1982): 20 ml of seawater arc ,filtered on Millipore HA filters, 13 mm

, diamcter; the filters are then kept dry and in darkness, and the measurements are made later. without extraction. The filters are stuck on a glnss plate and scanned in a Turner model I I I fluoromcter fitted with a thin-layer chromatography automatic scanning door. Thc fluorescence of the filters (F) allows the chlorophyll concentration (c) to be estimated using c= PA, where K is a calibration coefficient to be adjustcd periodically. Thanks to the simplicity of this technique, the crews of 10 to 15 voluntary observing ships who call at Noumda kindly carry out surface seawater filtrations at regular intervals throughout their route. making possiblc a continuous survey of the ocean. The cdunterpart of this simplicity is the error, which was estimated as 60% (DANDONNEAU, 1982); this high value can be iniputed to the long storage time at ambient temperature, to differences among species in the shape and size of the chloroplasts, and to microdistribution errors since thc filtered volume is only 20 mi.

6- ,

DATA A N A L Y S I S

The description of small-scale variations of biological parameters can often be carried out during a single experiment with a suitable’sampling scheme. When one is confronted with large-scale variations, there is often a problem of g5ps in the data (see for instance CHELTON er al,, 1982). Estimating a parameter’s value at a p!ace where there are no measurements, or defining how a measurement made at a place s must be taken into account for an estimation at a place x is a common problem when mapping results. A solution exists through the techni- ques Qf objective analysis which have been proposed by BRETHERTON el al, (1976) and K A R W E I T (1 980): assuming that ench observation may be decomposed into the form 8 = 4 + E (c= uncorrelnted error). the best estimate at a point x is given by

1

t

where N is the number of observations I). C,, is the covariance bctween Ox and 4). (a function of the distance b&ween .Y and r). and A,, is the matrix of covariance between all pairs or observations. The covariance between 4, and +s is C,; if r= s (corresponding to the diagonal of 1Iie matrix), the error variance of the observation must be added. The assumption that the expectation of the ficld be = O must receive particular attention here sincc the covariance

-.function is not known and must be computed using the same data set: estimating the cov.äriancc function is difticult if a significant trend remains.

The distribution of data is shown in Fig. 2. The chlorophyll concentration has a log-normal distribution, which is frequently observed in biological oceanography (FRONTIER, I973), nnd

. i r * ‘

$

I

I

I

3 , .. I

5

1380 Y. DANDONNEAU and F. GOHIN

.. .

1 1

tb)

( c l , i

' O L d O

20-

-2 O 2

1 Fig. 2. Frcqucncy distribution or the pata: (a) SSCC values: (b) X = log (I + 100 SSCC): (c) Y = X -f(la~itude, t ime);j is given by equation (3).

implies that the variance is dependent upon the mean, Thus, the error. was expressed in percent by DANDONNEAU (1982); the sampling error also depends upon the mean. A log- transformation IX = log ( I -t SSCC), where SSCC is expressed in mg 100 m-'1 normalizes the distribution of the data (Fig. 2) and stabilizes the variance.

The mean large-scale variations are unknown in this region, Our five-year SSCC series has been distributed into a grid (one month, 2 O latitude; the tíeglect of a possible zonal cffcct will be justified later), and then the results for the five years have bien averaged 'to produce the mean monthly values per 2 O latitude bands (Table 1). May to August appears as n rich

Tirble I. Meait vahies of log (I -+ SSCC) from January 1978 to December 1982, (Averaging procedirre: log

LI

Y

rrmrflormarloit; ineati for one monlh: niean for lheflve years. Data from lhe SUR TROPAC progrornnrcl

2 . J F M A M J J A ' S O N D

14-lh'S 1,007 1.760 1.944 1.597 2.306 ' 1.865 1.591 1:902 1.879 1,644 1.492 1.776 Ih-IX'S 1.942 2,480 2.443 1.456 1.974 2.291 2.044 '1.705 1,892 1.563 1.801 1.980 IX-20's 1.789 2.404 2.036 2.339 2.111 2.170 2.382 2.137 1.907 1.61 I 1.661 1.935 20-22's 2.007 1.947 1,881 1.712 2.163 2.343 2,436 2.074 1,983 1,632 1.716 2.132 22-24's 2.175 2.213 2.503 2.125 2.502 2.529 2.480 2.519 2.226 1.959 1,943 2.017 24-26's 2.179 2.309 2.041 1.977 1.978 2.728 2.922 2.798 2.740 2.207 1-85 I 1.958 2h-28"s 2.152 1.956 2.409 2.161 2.209 2.720 3.231 3.052 2.729 2,291 1.947 2.061 211-30°S 2,257 1.820 2.462 1,893 3.141 2,707 3.291 2.792 2.672 2.505 2.003 1.839 30-32OS 2.301 2.030 2.393 2.222 3.W4 3.008 2.621 3.209 2.650 2.392 2.207 2,155

,

4 6

. . - - 9

‘ir

I ‘-* SSCC in the tropical Pacific / 1381

period, mostly in the southern part: relatively high values also occur in February and March (austral summer) between 16 and 20OS. Subtracting these means from the individual data would produce an ensemble of values (monthly anomalies) satisfying 8, = O. However, the means listed in Table I which have been computed over five years are poorly defined (see for instance the high isolated value in May between 14 and 1 6 O 5 , or the relatively low one in April between 28 and 3 O O S ) . The following empirical prbcedure to express the mean annual variations has been favoured.

For each of the nine rows of Table I, corresponding to latitude L, the first three terms of the Fourier’s series decomposition have been computed; the mean variations at latitude L are then expressed as a function of time (I, in months) by ,

r .

The coemcients A I , L , . , , have been estimated by a polynomial function of latitude,’ . . . , B,, 1

using polynomial regressions of the first order for M, and of the fourth order for A The results are listed in Table 2.

The mean variations of log(1 +- SSCC) can now be computed using * *

Y(/, I) (Fig. 3) has been subtracted from each of the 1Qg (I‘+ SSCCJ, accounting for only

x, = log (SSCC,) - Y(/ , , I,) (4 ) corresponds to the instrumental and sampling errors, and to year-to-year variations. A third possible source of variance is the longitude. The residuals, X,, have been distributed into 15 classes (1’ longitude each) and the variance has been analyzed: the SNEDECOR F test

squares: furthermore, the major part of the data comes from the diagonal of the studied area (Fig. I ) , so that the longitude and latitude effects can hardly be dissociated. Hence the

I

I 16% of the variance. The remaining variance of

I

I

I

I

indicates a probability of 5,6 IOd for the null hypothesis, showing a zonal effect. However, the distribution of the data into 15 longitude classes only cuts out 1.5% of the total sum of

l ‘1.

longitude effect will not be taken into account, The XI residuals have been distributed into a one month by 2 O latitude and 15’ longilude

grid, Each grid point contains about six observations (3369 data for 9 latitude classes x I longitude class x 60 months) allowing k reasonably accurate estimate of the mean and random mean square of the X,, for. the latitude class I and the month I:

I

where N1, is the number of observations at the grid point (I, I ) . NI., varies from zero in poorly sampled periods and areas, to 20 on some occasions. The objective analysis will be carried out on the ,yla, which correspond to

... in equation (1).

j

I .

1

i ,

L

. .

Table 2. Results of the Fourier ana!vsis mrrietì ouî on [he data from Table 1. and cueficients used IO esrimale the meau r3ariulions o/ log (1 i- SSCC) according to equation (3)

Functions of latitude Correlation cmficicnt 14-16"s 16-18"s 18-20"S 20-22OS 22-24"s 24-26OS 26-28"s 28-30"s 30-37"s (1 varying from -14 to -32")

M A,

BI

*.l

variancc I s t harmonic 'h~variancc 2nd h;irmonic '!warinncc 3rd harmonic

1.805 1.964 2.040 2.002 2.266 2.307 2.410 2.449 2.523 M = 1.162-0.OJSI 0.98 0.022 -0.041 0.060 0.083 0.096 0.187 0.241 0.275 0.221 A, =-0.1694-0.13721 0.97

-0.018012-0.03083 1'

-0.057 -0.083

0.008 0.095

0.003 0.034 .

-0.089 0.040

0.001 -0.099

17 2d

O 22

37 24

-0.123 -0.035 -0.083

0.057 0.123 0.052

0.044 -0.034 -0.034

0.023 4.010 0.002

b '

0.019 o.Oo0 : -om3

57 29 6 6

16 58 16

3 O O

0.049

O. 126

0.099

-0.007

-0.043

55

38

3

0.033 -0.012

0.098 O.CH9

- * \

0.087 0.004

0.041 -0.026

0.026 -0.042

71 66

21 2

3 2

-0.0003121'

i- 0.06i) 12 f 0.00148 1'

--qo601'-0.(M174I' - 0X)O I88 i'

-0.0541'-0.001771' - 0.00002 13 I'

- 0.192 Iz - O.CH56 I' - O.CKO0503 1'

0.013 B, = 9.784 + 1.832 1 'i- 0.125 I* + 0.W371 I' i- o . m o 4 I'

-0.017 B,=6.108+ 1.021 0.84

+ O . m l 2 9 I' 0.029 A, = -5.232 - 0.924 1 0.75

-0.012 B, = -3.543 - 0.722 I 0.66

-0.068 A, = -16.009 - 2.59 I 0.9 I

0.49

-.\

69

I

7

-2

I a

1383 SSCC in the tropical Pacific

t '

Fig, 3. Standard year as given by equation (3) with thecocnicicnls M(!) , t f , ( I). and BI (I) as Iistcd in Tuble 2 .

The error is assumed to be constant oyer the field. I t has been estimated hereafter as the mean of th¿ a:, from the grid points (I, I>, weighted by d m * :

in which ulf given by ( 6 ) only accounts for. the subgrid noise. The summation is carried out only for the grid points where NISI is >3 and gives

Û' = 0.472. This subgrid noise corresponds to a variation coeflcient of 78%) for SSCC, al'tcr return to antilog, This value is high but is not surprising for a biological parameter at such a time and space scale (PLAT-F and DENMAN, 1975). The error for each x,., is then

Üfl = 6'/NI,l. (8)

The covariance function has been estimated from the In isotropic ficlds, the covariance is estimated as a function of only the distance between two points. The field we are coiicerncd with cnnnot be reduced to an isotropic one since time and space are heterogeqous dimcnsions. The form

where A/ and AI are the changes in latitude'nnd time, has been adopted, as it accounted besi for the observed covariance (Fig. 4). A , B, and C are constants which are computed ÍIS

follows: the mean value of all possible ,?I*, x z/+Al,,+A, is computed for A / = O', 2 O , . . . , 12" and AI = O, 1.2, , , , , 6 months, A special process is applied for A/ = AI = O: X I , is n n

C(A1, AI);= ABAICA1, (9)

* Tlis obscrved varianccs have been wcighted by d m rethcr than by2NI,, - I Tor the following reasws: ail is Jistribulcd as (a2/NI., - I ) x ~ ( N , , ~ -4): [he maximum likelihood of x ( N I , , - I\ and of the variancc of x (NI., - II lire respectively N,,, - I and 2(N/. , - 11. I t rollows that lhc variance oro/., is ,

a' x NI,, - I) 2a4 - --. (N/., -.Il2 4 1 - '

! t

1384 ' Y. DANDONNEAU nnd F. GOHIN

* A t ( m o n t h s ) I Fig. 4. Mcnn observed covnrinncct for lags = O I . . . ,6 months. O, 2 .4 , . . , , 1 2 O latitude, and ndoptcd covarinnce funcIion: C(w) = 0.1083 c-'.d585k* with )v = AI .I. 6.569 Ai, The units Tor AI nnd Ai hnvc bccn chosen in such I wny that w is thc rium of thc scgments AI and Ai. Ench point is thc nicrin value of nll possible ~ , , t A ' l t ~ , , + ~ for B givcn (LU. Ai), exccpi 'thc oint on ille Y.nxis

corresponding to A1 = At = O and whose ordinntc is J2 - b. estimate of the field thus we can write /

and

/

21.1 = & + &/.I

The left term of the sum is the true covariance; the last term is generally null since the error is uncorrelated over the field, but when Al =,Af = O the equation becomes: '

Thus the mean square error (7) has been subtracted from X i f io estimate C(O.0). A continual approach technique (in which the successive values of A , 8, and C are incre-

mented according to the second partial derivatives of (AI, AI) for A , B , and C) is used to minimize the sum of squares

1,

The successive results converge at A = 0;1083, B = 0.9434, and C = 0.6820, when AI is expressed in degrees and Al in months.

Equation (9) can be written in the form: log C(AI, At) = log A + AI log B i- Al log C, or log C(w) = log A 3- w log B, where w = AI + (log Ellog C)Ar and C(w) = A eb'", where 15 = log E.

,

The final form is then:

( 9 4 C(w) = 0.1083 e-O-0$n31''

RESULTS

With the covariance function (9a) and the error (7), we can optimally interpolate the ,Yl,, accordingto ( I ) . To save computer time and memory space, the estimates have been obtained using only the neighbouring grid points, at a distance such that the covariance function is >0.3 x C(0): since the Covariance function decreases more quickly with time (in months) than with latitude (in degrees), it roughly corresponds to the grid points over the whole latitude range, and from two months before to two months later.

). ... 5 .

I ,

I ' : . I

J

SSCC in the lropicnl Pacific 1385 I

The results are shown in Fig, 5a. The lowest values correspond to abnormally poor austral winters in 1978 and 1980 in the southern part, and to an abnormally poor summer in 1980-1981. The highest values correspond to rich winters in the southern part in 198 I and 1982, and to abnormally rich water during the summer of 1978-1979. The return to chloro- phyll concentrations (Fig, 5b) is made by: ( I ) adding Y(/, I) (equation 3) to the ,Y,,) estimates; (2) as recommended for log transformations of lognormally djstributed data, half of the logs variance is added to A',,, t Y(1, I) before computing the antilog and subtracting 1. The logs variance is 0.472, given by equation (7). The validity of this prbcess is checked by comparing the means of.the whole field: the mean value of Fig. 5b is 0.1 16, while the mean value of the 3369 initial data, which correspond to a non-uniform sampling effort, is O. 114. The most

, 'noteworthy 'ieaturt is the winter increase of the SSCC which spreads every year northwards to about 2OoS; this increase was weak in 1980, and especially in 1978. One can notice in the middle of the winter enrichment that the SSCC tends to be lower at 30 to 32OS than at 26 to 28OS during June and July. Relatively high summer SSCC are apparent in January and February 1978, in February and March 1979, during the whole summer 1979-1980, and in March 1982.

The objective analysis technique also provides the mean square error in the estimation of

Y

'

'

, ,

i

.. -

*. .

i

1386 Y. DANDONNEAU and F. GOHIN

the field: N

hl.f-1 u: = (e, - = C(0) - c c,,c,/f;:. (IO)

In log transformations X = log (Y), the variation coefficient of Y, is given by

Equations (IO) and (1 1) are used to produce Fig. !ïc, where the variation coeflÏcient of the SSCC estimates is shown. We can observe that the variation coefficient decreases from I978 where it is often >20% to the middle of 1981; it stabilizes after around 12%, corresponding to the best use of the voluntary observing ships network which provided about 75 datdmonth in the study area.

,..' D l S C U S S l O N

The phytoplankton grows in the whole euphotic layer, and the surface of the tropical sea generally exhibits a low chlorophyll content due to nutrient depletion. In the tropical Pacific. the subsurface chlorophyll maximum has been described as a continuous feature (VENRICK e! al.. 1973) in which is carried out a significant part of the primary production processes. and which is omitted by our sampling. Thus; before interpreting the variations of the SSCC, the relationships between the SSCC and Gthkr properties of the whole euphotic layer must be looked at. Using data from several oceanic arcas, LORENZEN (1970) has shown that the SSCC is significantly related to the chlorophyll content (r = 0.8 17) and primary productivity of the euphotic layer, but HAYWARD and VENRICK (I 982) found a less encouraging result, and even no corrclation in the central North Pacific. In the southwestern tropical Pacific, a similar study can be made using 150 oceanic stations from the same area (cruises from the Centre ORSTOM de Noumta, 1960 to 19130), giving log Chl O to 150 m = 6.53 + 0.557 log SSCC {Fig. 6). The correlation coefficient is smaller (r= 0.714) but still significant; the slope is not greatly direrent from that (b = 0.620) found by LORENZEN ( I 970) who used the chlorophyll content in only the photic layer, and whose data covered a wider range. As recommendcd by , HAYWARD and VENRICK ( I 982). cautioh is necessary before SSCC variations be considercd as variations in the whole photic layer.

The data from these 150 stations have been examined to identify the causes 01' the winter increase of the SSCC which appears in Table I and Fig. .fib. At each station, the maximum vertical temperature gradient, the depth of this maximum, the depth of the nutricline (defined here as the depth where NO, concentration becomes > I pmol/l) and the depth of the chloro- phyll maximum are considered in two cases: data from June to September south of 23OS corresponding to the winter increase, and data from north of 2I0S, where the winter increase does not appear, The first group of data exhibits smaller values of the maximum vertical gradient of temperature, and these maxi,ma are found at greater depths than in the northern part (Fig. 7a and b). Both in the north 6nd in the south, the maximum gradients are weak and these results support the hypothesis that the winter cooling of the sea surface (at Noumia, 22OS, the range of the mean annual cycle of temperature is about 6OC) favours vertical mixing. DANDONNEAU (1 979) related relatively high chlorophyll concentrations in this region to the vertical instability and thc potentially important island mass effects. Since the southern

'

'

100 -

50 -

\

20-

1 ° j , * , , . , ,

*. ,

* I .

0.05 0.10 0.20 0.50 1.00 sscc (mglm3 I

' tropical Pacific (dala from the oceanogrlaphic cruises of the Centre ORSTOM de Nouméa). Log (Chi O to 150 m) = 0.557 log SSCC + 6.53 (r = 0.7 I).

part is free from islands (Norfolk excepted), the rougher weather in wintcr is probably the main cause of vertical mixing. Thus, while the thermocline deepens (Fig. 7b), the nutricline l i b and nitrates are present at the surface at a significant number of stations (Fig. 7c); identically, the subsurface chlorophyll maximum rises and is often found in the mixed layer (Fig. 7d), Such conditions are similar to , those which prevail .during the, autumn,,bloom in temperate arcas rather than to typically tropical structures characterized by a nutrient depleted mixed layer (HERBLAND and VOITURI$Z, I979), Thus, the widely present'subsurface chlorophyll maximum in the Pacific Ocean (VENRICK et d., 1973) would not extend southward farther than about 22OS during winter in thc southwestern tropical Pacific; this limit corresponds to the position of the tropical convergence in August (ROTSCHI and LEMASSON, 1967). However, the winter conditions in the south for nitrate and chlorophyll

I (Fig, 7c and ci) appear extremely variable: in the northern part,:he depths of the thermocline (Fig. 7b) and of the nutricline (Fig. 7c) have unimodal distributions with means respectively at 134 and 129 m; in the south, two distinct situations occur, i.e., a deep nutricline, or nitrate present in the mixed layer, corresponding respectively to a deep or superficial chlorophyll maximum, ,,

In the central gyre,of the North Pacific at 28"N, MCGOWAN and HAYWARD (1978) observed q mixing regime that they believed was maintained by a series of mal1 mixing events: no change was observed in the phytoplankton standing crop but the primary produc- tion and the zooplankton biomass increased significantly, Recent.ln siiu IJC incubations at 23"s indicate that the carbon fixation rate is about 0.5 g m" d-' in winter in the studied area, corresponding approximately to a two-fold increase if compared to the summer values; the zooplankton populations in the vicinity of New Caledonia are different from those in the oligotrophic central Pacific at 15 to 20°S, and present some similarities with those influenced

' by the equatorial upwelling in the eastern Pacific (DESSIER, 1982) arguing for an active new

' I .,,production in the southwestern tropical Pacific. However, the mixing events here probably occur more frequently than in the northern Pacific at 28ON, since we observe a significant increase in the phytoplankton biomass (Fig. 5 and Table I).

< Fig. 6. Relationship between SSCC and integrated chlorophyll (O to I50 m) in the southwestern . ,

t ,

' .

!

'I ,

\

' I

., ,

(d 1

1388 Y. DANDONNEAU and F. GOHIN

TROPICAL CONDITIONS WINTER CONDITIONS

h O61 'l. 051 '01 IO . . b1 , (a )

O 5 10 I5.lIn OS I O 15Ym O

Drplh 01 l h e lhrrmotl inr

Orplh *i Ihr nult lc l inr \'

l

i Fig. 7. Frcqucncy distributions of some featurcs of vcrtical profiles in tropical (Icn) and winlcr (right) conditions. The tropical conditions satisfy: months I to 12, latitude 14 to 21 O S . longitude I60 to 175OE. Winter conditions: months 6 to 9.latitude23 to 3Z0S, longitudc 160 to 175'E (data from thc occnnogrnphic cruises of llic Centre ORSTOM dc Noumia). (a) Maximum vertical tcmperncurc grndicnl (computcd in depth intervals -10 to 20m). (b) Depth or this maximum. (c) First dcpLh

whcrc NO,- concentmtion is $1 pmol/l. (d) Depth or the chlorophyll maximum.

The mixing events are likely to occur during periods of strong winds when the sea surface has been cooled in winter. .The heat loss to the atmosphere increases south- wards and the increase is especially marked around 20°S from April to August; the mean loss in June is ISOWm" (WEARE el al,, 1980) at 24OS. The year- to-year variations which appear in Fig. 5b 'are examined in this respect. No significant differences have been found in the sea surface temperature variations between the five winters

C I

1

I

' ,

1978 10 1982 (except the warmer 1978 winter). The frequency of strong winds vnries bctwcen years and the winter winds (June to September) south of New Caledonia were stronger i n 1979, 198 I , and 1982 than in 1978 and 1980 (Fig. 8). The calm weather in 1978 and 1980 can thus explain the weakness of the winter increase during these years. A similar relationship was also noted by MENZEL and RYTHER (1961): it shows that the SSCC winter increase is the consequence of wind-induced vertical mixing, which thickens the mixed layer and erodes the nutriche. Comparing the depths of the nutricline (Fig. 7c) to the thickness of the mixcd laycr (Table 3) indicates that the order of magnitude of this erosion is several tens of meters, and that a large amount of nitrates is introduced into the mixed layer and made available to photosynthesis. The mean depth of the surface mixed layer approaches 200 m in July and August south of 28's; at such latitudes the decrease of the solar incidence and daylight length in June and July is enhanced, and while nitrates are present a t the surface, the phytoplankton growth is probably limited by the unfavourable ratio between the compensation depth and the

I

... . I

SSCC in the tropical Pacific 1389 I

n

Fig. 8. Frc uency of the observed wind speeds in the area limited by 22 and 32OS. 160 and 175'E. Upper linc: &d speeds >8 m 4-l (Beaufort forcc 25). Hatched: wind speeds > 10.8 m s-I (Beaurort

force 36). (Voluntary observing ships data of the SURTROPAC program")

/

mixed-layer thickness (SVERDRUP, 1953). This is a plausible explanation for the relatively IOW SSCC in winter at 3 0 to 32OS in June and July, and the SSCC data from voluntary observing ships make i t possible to identify 20 to 2 2 O S as the northern limit of the winter uni-moda! bloom, and 3OoS as that of the winter bi-modal bloom. A SSCC variation also appears in the northern part, with a maximum centered in February

(Table I and Fig. 3). This feature does not repeat as regularly as the winter increase and its importance here is due largely to the high SSCC value in February 1980 (Fig. 5b). Relatively high summer values also occurred in 1978, 1979, and 1982. They seem to result from blooms of the nitrogen fixing blue-green algae Trichodesmiunr spp. (= Oscillatoria) which are often reported in summer conditions by the oficers of the voluntary obserying ships. REVELANTE and GILMARTIN (1 982) observe that these algae are often dominant in the Great Barrier Reef Lagoon, close to our area, and connect their abundance to low wind periods. Around New Caledonia,, they have also frequently been observed during the cruises of the Centre ORSTOM de Noumta, spreading in long and dense yellow stripes at the sea surface. Due to their specific physiology, these algae tend to poncentrate at the surface where they assimilate rhc atmospheric nitrogen, but they are dispersed into the mixed layer when the wind streng- thens. The low wind speeds observed between 24 and 14"s are shown in Fig. 9, The calms which appear during [he first months of 1978, 1979, and 1980 correspond to relatively high summer SSCC. No such high SSCC occurred in early 1981 during which the frequency of calms was only slightly belo3 that of 1980; however, one can notice that the calms occurring in May 1981 correspond to high SSCC which coalesce with the 1981 winter bloom, and are more likely due to proliferations of Trichodesmiwn rather than to an abnormally early +inter bloom. Calms were scarce too in I982 (Figs 8 and 9 show a trend for the winds to increase in strength from I978 to 1982; it probably corresponds to the increase of the southern oscilla- tion index during the same period), but the calms maximum in March corresponds again to a SSCC increase. The effect of the wind is therefore probably determinant for both the summer Tric/iodesmfirtn bloom and the winter bloom, but this effect works in opposite directions

+

'

'

lbhle 3. Meait rhlrkness of rhe surface mixed layer (T > T siir/acc - I OC) arid s f c r r r h r d deviariori beweeit I55 arid 175 E. (XBTdarafront N O D O I . I

20-24's 24-285 2a-320~ , June IV m (30) 126 m (4) No dala July 93 m (58) 134 m (24) 198 m (56) August 126m(22) I69 m (42) I97 m (56) Scprc" her 60 in (24) 154 m (85) 156 m (67)

f'

IV

8 .

*. .

A.

I

.I 1

1390 Y. D A N W N N E A U nnd F. GOtlIrJ

50

O

I sia I979 lb80 i g a i 1982

Fip. Y, Frcqucncy of tlic observcd wind spccds in thc arca limited by 14 and 24's. 160 and I7S'E. Uppcr line: wind speeds (3.5 m s-I (Beaufort forcc (2). Hn1chcd: wind spccds < 1,s ni s-I

(Beaufort force Gl). (Voluntary obscrving ships,dala of the SURTROPAC programme)

according to the case: in winter (in the southern part) an increase of the wind speed produces an income of new nutrients into the mixed layer and causes an increase of the SSCC: in summer (and possibly all the year round in the northern part) the pycnocline cannot be eroded so intensively and an increase in the wind speed will more likely disperse the Triciroclesn~i~rru patches into the mixed layer, and so decrease the SSCC. With reference to the warning of I-I~vwnno and VENRICK (1982) for oligotrophic areas, it must be pointed out that thesc patches arc sea surface features, so that the resulting SSCC increase probably does not reflccl a similar increase in the whole water column.

The covariance function given by (9a) indicates that the SSCk varies more quickly in one month than in 2 O , allowing some comments to be made on the cause of the variations, I t has been shown abovc that the wind was an important factor in this region. In winter, the main meteorological events are depressions moving eastwards, ~ 5 0 0 km in diameter, and generally associated in series which affect a large part of'the Coral and Tasman seas: they do not usually last more than 10 days and are replaced by a large anticyclone. In summer, tropical storms move southwards through the whole,study area in a few days; they occur in periods

magnitude for the length of the events. ThÚs the size of the meteorological events. 5 l o 15 days, and 500 to 1500 km, roughly corresponds to lhe covariance function and so confirms the role of the wind variations in the dynamics of the phytoplankton in this region. This of course excludes the small-scale variations which are smoothed by the one month and 2 O in Iatitude, 1 5 O in longitude grid. This grid is cÓarse, since it does not consider features occurring inside an area 2: 300,004) km2. The subgrid yariance given by equation (7) is high, and i t has been shown that introducing a longitude grid only removed a negligible part of the variance, Thc island mass effects possibly represent another source of variance, at least north of 23OS: they imply, however, a complex field which could not be properly studied herc.

where calms and trade winds alternate, andvagain 10 days may be considered as an order of a -

Equation (3) can be used to predict SSCC at a given time and latitude: l.

where a2 = 0.472 is the subgrid error variance on the logs. The precision of the prediction can be evaluated using the ratio of the standard year range to the anomaly. Assuming that

,SSCC = S W C + r, the standard year SSCC is given by (12); the anomaly r is obtained by subtracting (12) from the results mapped in Fig. 5b. The ratio of thc standard d¿viations of SSCC and r varies from 0.5 at 1 5 O S to 1.1 1 at 29OS (Fig. IO). Consequently, cquation (3) poorly predicts SSCC in this region since the unresolved term is of the same order of mag- nitude as the determined one (and twice as large at low lalitudes).

rcv

w

c '

SSCC in the tropicnl Pacific

mean

1391

!

! I

Seasonal variations of SSCC and their limits in latitude have been shown in a wide tropical area, These variations d o not represent a major part of the variance, and che temporal varinbility is high, so that their 'description requires observations at short-term intervals which can hardly bc carried out between 14 and 32OS using oceanographic cruises. Voluntary observing ships which have been shown to be helprul in meteorology, in physical oceanogra-

lphy. and even in biological oceanography (HARDY, 1939), prove to be reliabjc for phytoplankton studies. Large-scale studies in primary productivity need the use of indices, sincc the most common method to measure photosynthesis-the "C technique-cannat be performed with a high enough output. SSCC appears here to be an interesting index if eriough data allow the Small-scale noise to be smoothed. Further developments in satellite techniques will probably make SSCC the most easily obtainable index. Its use however requires a pre- vious knowledge of the biological properties of the water masses. The winter increase in SSCC in the southwestern tropical Pacific corresponds to an increas'e in primary production unlike the irregular summer increase which corresponds to 7Wchhodesmium spp. blooms, These blooms only affect the sea surface, and the responsible species,have a very low produc- rion rale: nevertheless, they represent an entry of nitrogen into the food chain and constitute a unique ccosystem (CARPENTER, 1973) which is probably latent ad summer in this region, developing during calms.

The amplitude of the seasonal variations is much lower here than in temperate or inter. mittent upwelling"z.ones. Due to the Considerable small-scale variability, the seasonal cycle

,~ must be ascertained on a statistical basis. The winter increase south of 20 to 22OS seems to be

regularity; recent data from the SURTROPAC programme indicate that in 1983 the winter ' ' bloom started in June and was well developed, and that a slight.summer increase occurred

from February to April between 19 and 23OS, thus confirming the proposed scheme,

I an unquestionable feature: the summer Trichodesmlum spp. bloom does not show as much

,

,

" I " t . . . .

,..<,*.i

' 4.,

J

\ I

I392 Y. L ~ N D O N N E A U and F. GOHIN

' Arkiroccik~~orreir/s-Wc arc indebted to the crews of the merchant ships who arc kindly and rcgulnrly taking Ihr chlorophyll samplcs in the whole tropical Pacific Ocean. Jean Paul Rebcrt was a prudent adviscr in thc nnnlysis 01 thc dntn. Wc wish also IO thank Alain Morliirc for managing the SURTROPAC data and providing dccisive hclll in computing. Henri Walico for his patient watching over the mcasurcmcnls and sampling cquipmcni, and l c n n Rene Donguy for his dctcrmining pnri in the working of the programme.

REFERENCES

BEERS J . R.. D. M. STEVEN and J. B. LEWIS (1968) Primary productivity in the Carribcan' Sea ofTJamaic,i and t h t tropical north Atlantic OTT Barbados. Biil/e/in of Marine Science, 18. 86-104.

UIENFANG I'. K. nnd J. P. SZVPER (1981) Phytoplankton dynamics in the subtropical Pacific Ocean OTT H a d i DetpSen Research. 28. 98I-IWO.

Udonov V, G. (1958) PcrspLTtivcs in the study of seasonal changes of plankton nnd of the number of gcncrailonL ni difkrcnt latitudes. In: i'erspcc/lves in niarine biolos)', A, A. BUZZATI-T~AYERSO. editor. Utiivcrsit!' 11

California Press, Bcrkclcy. CA, pp. 145-158. URETITHERION F. P., R. E. DAVIS and c. &, FANDRY (1976) A tcchniqU,C for objcctivc analysis and design 01

occnnugrnpliic cxperimcntv npplicd to MODE-73. Deep-sea IjCsearrli, 23, 559-582. CARPENTER E. J. (1973) Nitrogcn fixation by Orcl//alorla (Tric/iodesnr/itm) thiebaurii in the southwestcrr.

Sargasso Sea. Deep-Sta Research, 20. 205-288. CHELTON D. E.. P. A. BERNAL and J. A. MCGOWAN (1982) Large-scale interannual physical and biological

internclion in the Catirornia currcnt. Jortrnal of Marirre Research. 40, 1095-1 125. CUSHING D. H, (1959) Thc rensonnl vnrintion in oceanic production as a problem in populations dynamic$

Joirnrnl dit Conseil, Conseil Permaneril International p o w I'ExploraIlon de la Mer, 2 4 , 455-4 64. DANDONNEAU Y. (1 979) Concentrations en chlorophylle dans le Pacifique lropicnl sud-ouest: comparaison avec

d'autres nires ockaniquer tropicnlee Oceanologica A a a . 2, 133- 142. . DANDONNEAU Y, (I 982) A method ror the .ripid determination of chlorophyll plus phacopigmcnts in samplcs

collected by merchant ships. Deep-sea' Research. 29. 647-654. DESSlER A. ( I 982) Variabiliti spatiale el snis6nniirc des peuplements &piplancloniques des copepodcs du Pacifiquc

tropical sud et iquntorial (est Pacifique). Oceorrologlca Acto. 6, 89-1003. F R O N T I E R S, (I 973) Etude statistique de la dispersion du zooplancton. Journal o/ E.vperirnenra1 Biolop om/

HARDY A. C, ( I 939) Ecological investigations with the continuous plankton rccorder: object. plan and methods. Hi111 ßiiIle11~1 of Marine Ecolog): I, 1-57,

HAYWARD T. L. nnd E. L. VENRICK (1902) Relation bctwcen surhcc chlorophyll and integrated primary produc t ion . M a rine Blologj: 6 9, 2 4 7-2 5 2.

t-1ERDLAND A. and ß. VOITURIEZ (1979) Hydrological structure analysis for estimating the primary production ¡ir

the tropical Atlantic.Joirrna/of Marine Researcir. 37.87-101. KARWEIT M. (1980) Optimal objective mapping: a technique for fitting surfaces to scattired data. In: Adiwrrrti

coircep/s III orean measuremenlsJor p v h e bfolog~: F. DIEMER, J. VERNBERG and D. ~ V ~ I R K E S , cditorr, Univcrsity of South Carolina Press, Columbia. pp. 8 1-99.

LORENZEN C. J. (1970) Surface chlorophyll ns nn index or the depth, chlorophyll content, and primary productivity of the euphotic layer. Llfiriology and Oeeonogruphy. 15, 479-480.

MCGOWAN J. A. and T. L. HAYWARD (1978) Mixing and occnnic produc[ivity. Deep-sea Resmrch. 25.77 1-793. MENZEL D. W. and J. H. RYTHER (1961) Annual variations in primary production of tlic Sargasso Sca OTT

Bermuda, Deep-sea Research. 7, 282-288, MINAS H.J. (1970) Ln distribution de I'oxyginc en relation avec la produbion primaire en Médilcrranec Nord

Occidentale. Marine Biology, 7. 181-204. . OWEN R. W. and B. ZEITSCHELL (1970) Phytoplnnkton production: sensonal change in the oceanic eastern

tropical Pacific. Marlne B/o/ogY, 7, 32-36. PLAm T. and K. L. DENMAN (1975) A general quntion for the mesoscale distribution of phytoplankton in thc

sen. Mémoires de la SociCf6 Royale des Sciences de Liége, 6th serles, I . 31-42 . REVELANTE N. and M. OILMARTIN (1982) Dynamicsof phytoplankton in thc Great Bnrricr Rccf Lagoon. Joirntnl

of Pln~rktorr Research, 4. 47-76. ROTSCHI H. and L, LEMASSON (1967) Oceanography of the Coral and Tasman seas. Oceanography a d rtforim,

Biologi* Anniru1 Review. S. 49-97. SCHULENDERGER E. and J. L. REID (1981) The Pacific shallow oxygen maximum, deep chlorophyll maximuni.

and primary productivity, reconsidered. Deep-sea Research, 28. 90 1-9 19. SOURNIA A. (I 969) Cycle annuel du phytoplankton CI dc la production primaire dans Ics mers tropicalcs. rtlnriuc

SVKRDIWP H. U. (1953) On conditions for the vernal blooming of phytoplaiikion. Joirrrral dit Coriseil, Corisei( Per.

1

ECdoKi: 1Z. 229-262.

!

I l l~ l~g i* , 3, 287-303.

ninirelit i~iternationd pour /'Explorailon de la Mer, 18. 287-295.

/ '

.-.

l . I ., .

SSCC in the tropical Pacific 1393

VENRICK E. L.. J. A. MCGOWAN and A. W. MANTYLA (1973) Dcep maxima of photosynthetic chlorophyll in the

VOITUREZ B. nnd A. HEMDLAND ( I 98 I) Primary production in the tropical Atlantic ocean mapped from oxygen

WEARE B. C.. P. T. STRUB and M. D. SAMUEL (1 980) Mnrinc climate Allas or the tropical Pacific ocean. Coir-

Pacific ocean. Fishety Birllc~ìn, 7 1, 4 1-52.

values of Equnlant 1 and 2 (1963). B u l b h i of Murine Science, 31, 853-863.

Irlbulloits AIiJlosp/fertc Scieirce, Univcrsily of Calirornin, Davis, No. 20. I47 pp.

I

f'